WDG Journal
Reporting on water data and the machine learning built on it.
Why one dataset of 6,830 catchments matters more to flood AI than any new model
Caravan stitched regional streamflow collections into one standard format. For machine-learning hydrology, that is the bottleneck it removed.
Earlier this week
Global flood forecasts now run on an AI weather model. What changes for users?
GloFAS v4.4 moved its operational forecasts onto ECMWF’s AIFS. The flood model downstream is the same; the weather it is fed is not.
Fourteen million wells: the map of where America actually pumps
A 2024 database compiled state well records into one national picture of depth, location and construction. Drought models have rarely had this much detail.
A century of river chemistry, in one archive
GRQA gathers 17 million water-quality observations going back to 1898. Paired with a global map of treatment plants, it becomes a training set for pollution models.